In a typical image search, a user specifies one or more search terms in order to locate images corresponding to the search terms. An image typically has metadata associated with it. An image search generally works by examining the associated metadata to determine if the metadata match a user-specified search term. Typically, images for which the associated metadata match a user-specified search term are then returned to the user.
One problem that a typical image search poses is that an image is unlikely to have associated metadata that adequately describe the image. It has been said that a picture is worth a thousand words. However, it is unlikely that an image has such a large amount of associated metadata. Even if an image has such a large amount of associated metadata, it is unlikely that the individual elements of the associated metadata would combine in a coherent manner to evoke the same response in a user that viewing the image may evoke.
Accordingly, a system that allows a user to search for images by a method that improves upon a typical image search would have significant utility.